FIRST, DO NO HARM!

For a good decision it’s necessary to be well informed. Therefore I will lead you through some mindsteps in this document which I really find important. Even though this report is about germany you can take this process as a guideline for your own country. At first an individual has to assess the risk of a disease (in this case Covid). Not by media and how big the numbers are or how deeply

!!!RED!!!

the background is priming and framing you with pictures from hundreds of dead people but about rational calculations. As a second step, which is the primary focus in this document, someone has to assess the risk for Adverse Events (AE) especially Severe Adverse Events (SAE) from the medicine (in this case the Covid-Injections). A third step is to weigh the risks against each other to get a benefit-cost-analysis. It is crucial that the individual cohorts are taken into account. There is no “one fits all” solution. One dimensional thinking isn’t appropriate to tackle such a complex problem. Clustering by age alone also can’t fix this problem (https://twitter.com/Thomas_Wilckens/status/1491762388617551876). Even after clustering the “right” vulnerable group a decision must always be well informed and free. First, do no harm!

Risk of Disease

The Population in Germany counts 83.2 million people in 2021. Nearly half of that are womean (50,7%) and men (49,3%) (https://www.destatis.de/DE/Themen/Gesellschaft-Umwelt/Bevoelkerung/Bevoelkerungsstand/_inhalt.html). Over approximately 2 years there were 12,391,463 recognized cases of Covid of which 119,939 died as of 14.02.2022 (https://coronavirus.jhu.edu/data/mortality). Sadly we still don’t know in how many cases Covid was the reason. Maybe we will know in the future (https://www.bild.de/politik/inland/politik-inland/corona-patienten-sollen-endlich-richtig-gezaehlt-werden-neue-datenerhebung-gepla-79099740.bild.html). Acutally we must claim that not all death were due to Covid (https://www.berliner-zeitung.de/news/falsche-daten-viele-corona-patienten-liegen-nicht-wegen-corona-im-krankenhaus-li.208423). But if they all were due to Covid the Case-Fatality Rate is 0.97%. This means that, the risk of an individual to get Covid is (historically) 14.89% and to die from Covid is 0.14%. Digging deeper shows the distribution by age and sex.

Death_Covid_by_Age_and_Sex


Chance of dying with/because of Covid by age group and alternative death ratios:

Death_Covid_by_Age_and_Sex2

Note: Marginal diff. in cases and death to J. Hopkins because of little diff. in Time.

Median and mean Age for Death in correlation with Covid and Historie:

Death_Covid_by_Age_and_Sex2

Long Covid:

An argument which often occurs in this context is long Covid. “I don’t fear to die, I fear to get long Covid”. In this document I will not dig deep into this topic. I think there are a lot of good studies out there.

I just want to share the next illustration. For me it seems like LONG COVID is another instrumentalized topic:

Long Covid Kids: Death_Covid_by_Age_and_Sex2


Risk of Adverse Events

Pfizer

Any medical interventions must first be proven safe. “The vaccines are safe and effective”. Firstly, Pfizer’s own documents don’t seem to conclude that.

Canadian Covid Care Alliance


Efficacy of the mRNA-1273 SARS-CoV-2 Vaccine at Completion of Blinded Phase Efficacy of the mRNA-1273 SARS-CoV-2 Vaccine at Completion of Blinded Phase

VAERS

The Paul-Ehrlich-Institut (PEI) - (https://www.pei.de/DE/home/home-node.html) transfers it’s reports to (foreign) VAERS, which you can see in the (SYMPTOM_TEXT). At least 85% of all records in foreign VAERS for Germany are from PEI. These reports are the fundament of this analysis.

If you look at the chart above you can see that 2021 is a really big outlier and even 2022 is in mid-february already as high as the expected amount of 2021. Now we have to ask two questions.

  1. Are there really more adverse events in 2021 and 2022 than in 2020 or are they just in this passive surveillance system?
  2. Are there more reports in this system because of the high attention in the public and the higher share of fearful persons?

Even though the passive surveillance systems are highly underreported, as usual in every year, it’s plausible that in this range of underreporting there could be really an increase in reporting for Covid-Injections in the last 2 years. So only a simple comparision over time is not really meaningful. But if there is more attention and more fearful persons who reported Adverse Events we would expect that this increase would fluctuates in ranges over all Covid batches.

In the chart above you can see extreme differences in reports relating to Covid batches, between manufacturer and within manufacturer. That’s not what we would expect relating to our first 2 questions. We would expect an increase in reports but a variability which is seen in all batches like the flu-vaccine for example (https://www.bitchute.com/video/8wJYP2NpGwN2/)

Now there could be another reason for this extreme differnces in variability.

  1. Are there a lot of typos in the Database(VAERS)? After correcting for typos, is there a more stable and expected variablilty picture?

Note: All AstraZeneca batches have equal or less than 5 records.

After correcting for typos the picture got a little bit clearer but it keeps the pattern of extreme differences between batches and manufacturer. Before digging deeper into this subject here is the plot for Severe Adverse Events (SAE).

The pattern is pretty the same. This is what must be exptected based on the AE charts. More AE leads to more SAE in absolute numbers (more dosis more damage). To get this into perspective we need now the batch size from the manufacturer to measure ratios. Sadly we don’t have valid data for all batches in history. But what we can measure is the ratio of toxicity. This means we divide SAE by AE per batch/lot. Independent from batch size we expect a stable toxicity range for all batches. The Next Table shows toxicity ranges for 4 different cases (AE > 10, AE > 25, AE > 50 and AE > 100).

##      TOXICITY_10         TOXICITY_25         TOXICITY_50      
##     "Min.   :0.03846  " "Min.   :0.03846  " "Min.   :0.2143  "
##     "1st Qu.:0.51485  " "1st Qu.:0.51282  " "1st Qu.:0.5258  "
##     "Median :0.58261  " "Median :0.57627  " "Median :0.5719  "
##     "Mean   :0.57405  " "Mean   :0.56309  " "Mean   :0.5636  "
##     "3rd Qu.:0.65657  " "3rd Qu.:0.63399  " "3rd Qu.:0.6030  "
##     "Max.   :0.94444  " "Max.   :0.82857  " "Max.   :0.7544  "
## IQR "0.14171"           "0.12117"           "0.07722"         
##      TOXICITY_100     
##     "Min.   :0.3488  "
##     "1st Qu.:0.5203  "
##     "Median :0.5576  "
##     "Mean   :0.5504  "
##     "3rd Qu.:0.5828  "
##     "Max.   :0.6964  "
## IQR "0.06245"

Note: Severity is calculated as: If a person was in hospital, had life threatening, disability, was in emergency room or died. If it matches one criteria it is one severe case.

As you can see in the table above. Batches with < 25 records are, as expected, more sensitive because of the base effect. Here is the plot for severe cases from batches with > 50 records.

That looks pretty stable right? But at first lets make the Y-Axes more clearer:

There are Batches with ~35-47% toxicity and there are batches with ~55-75%. This means a range of 35% to 75% so a difference of up to approximately ~40%. So lets calculate the p-value for the difference in batches between Pfizer and Moderna (I don’t choose Janssen because there is only 1 report per batch > 50 - XE395) based on the first toxicity measure.

We can see in the chart above that there seems to be no significant difference in toxicity (based on this measure) between Pfizer and Moderna. Later in this report I look within the same manufacturer UPDATED: Within manufacturer (Pfizer)).

We can’t reject that batches between Moderna and Pfizer are significantly different based on toxicity ratio right now. But later in this report we have to reject that batches from Pfizer and Moderna are equally within the manufacturer. This reminds me a little bit of (https://en.wikipedia.org/wiki/Simpson%27s_paradox).

Second 56% of all records are severe. This could be because many people report Adverse Events if they are not very mild. For example headache for some hours or short fever or things like that is probably less reported. In reality probably many people just report things when they feel very uncomfortable.

Third by only looking at how many reports are severe we can’t difference how much more severe one case is compared to another. Therefore we have to summarize the multiple severe outcomes (e.g. life threatening, disability and following death is more severe than “only” hospitalized) to compare the level of severity between batches.

Therefore the next 2 Measures are (https://howbad.info/lethal.html):

  1. Lethality = reports that are fatal = deaths/total number of reports
  2. Severity (2.0) = reports that are multiple severe = (deaths + disabilities + life threatening events)/total number of reports

First I will cut out all batches with AE <= 175 to get a more stable view and to avoid high severity and lethality because of batches with low counts of reports (base effect).

Now the chart above isn’t even looking stable any longer. It seems like there are strong differences in severity. Not only there are differences in severity by a factor of ~3 but also differences in lethality by a factor of approximately 5 to 7. Another thing which gets obvious is that some batches are so severe that they are highly unusual in the Population (red line) based on a Bootstrap-Distribution.

Note: The CI in the chart above is calculated with AE > 175 to be very good comparable.

Here is the chart for lethality:

Bootstrap CI

What is the 95% and 99.9% CI for severity in the Population (Germany)

Here you can see the distribution in the Population based on Bootstrap (10,000x) CI with AE > 10.

The 95% and 99.9% CI is therefore:

##        5%       95% 
## 0.1301583 0.1578287
##      0.1%     99.9% 
## 0.1193246 0.1722045

And the distribution of (mutiple) severity ratio from the sample size is:

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##  0.0000  0.0750  0.1206  0.1434  0.1824  0.7273

What does that mean? It means that batches with < 12.8% and > 16.1% are very unusual in the Population. Now imagine where batches with 0.08 or 0.2 are in the distribution.

Hint:

Note: The CI is calculated with AE > 10. This is very conservative because of the base effect and the high severity rate in batches with low AE. Calculating with AE > 100 make this picture dramatically worse! E.g the upper 99.9% CI for AE > 175 is: 0.147 (upper red dotted line in the Toxicity 2.0 chart).

Here is the Bootstrap-Distrubtion for Lethality (AE > 175):

##         5%        95% 
## 0.03748825 0.05435764
##       0.1%      99.9% 
## 0.03184784 0.06330036

Is the Effect (multiple severity) significant?

Between manufacturer

Now we want to know if we just maybe have problems with our eyes and the effect is strongly visible but maybe we have cognitive biases like Müller-Lyer-Illusion or so. Therefore In this chapter I will analyse the significance by testing with permutation. In this chapter I will compare the batches between the three manufacturer (Janssen, Moderna & Pfizer). I will make three permutation tests to test the significance between the 3 pairs (Janssen vs. Moderna, Jannsen vs. Pfizer and Pfizer vs. Moderna).

As we can see in the chart above, even if the distribution is a bit skewed, we can’t reject the H0 (Jannsen & Moderna are equally severe) by the standard significance level (0.95, 0.99, 0.999). Nevertheless as you can see the blue dotted line (difference in mean by these groups) is also not really near the 0. It is more near the 0.9 significance level. Keep in mind from earlier explanation that 2 groups can reverse an effect within groups.

The comparison between Janssen & Pfizer is really another picture. The chart is a little bit skewed but that is no huge problem. The p-value is highly significant. So here we must reject the H0 (Jannsen & Pfizer are equally (multiple) severe).

Also the comparison between Pfizer and Moderna is highly significant with a p-value of 0.0155. This is really interesting because in the single severity measure there was no significant effect. It seemed like they canceled the effect each other out. Therefore it’s necessary that we will at least drill down to one manufacturer again which happens in the next step.

At this point we have to say that it seems like Pfizer is significantly different in severity from the other manufacturer. This is what we also saw in the Toxicity 2.0 chart. It seems like the following order appears: Janssen < Moderna < Pfizer. But there is something we have to remember. The number of reports are also Jannsen < Moderna < Pfizer. Pfizer was in germany the most dominant injection. Here is the table of distinct batches by manufacturer > 10 AE.

## # A tibble: 3 x 2
##   VAX_MANU           Count
##   <chr>              <int>
## 1 "PFIZER\\BIONTECH"    99
## 2 "MODERNA"             34
## 3 "JANSSEN"             12

So to get rid of these 2 potentially biases the next step is looking at just Pfizer batches alone.

UPDATED: Within manufacturer (Pfizer)

I’m very sorry for the first Version. I calculated the p-value by dividing the Pfizer-Batches into two groups based on the mean. The problem with this process is that you literally get always significant p-values around 0 even if differences between groups are completely marginal. So this process was not suitable to tackle this problem. At least this is the case for two sided tests. I changed the process so that now I compare the differences between the Pfizer series as also 4 single batches from the “E” series.

At first I calculated the analysis of variance (ANOVA) for the Series with more than 11 AE:

##             Df Sum Sq Mean Sq F value   Pr(>F)    
## Series       4 0.1525 0.03812   8.355 8.72e-06 ***
## Residuals   91 0.4152 0.00456                     
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

As you can see the p-value is highly significant which means that we have to reject the Null-hypothesis that there is no difference between these series.

(ANOVA) for the Series with more than 50 AE:

##             Df  Sum Sq  Mean Sq F value Pr(>F)  
## Series       3 0.01958 0.006527   3.198 0.0299 *
## Residuals   58 0.11838 0.002041                 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

As before we have to assume that these series are not the same for severity.

(ANOVA) for the Series with more than 175 AE:

##             Df  Sum Sq  Mean Sq F value Pr(>F)
## Series       2 0.00381 0.001904   1.245  0.302
## Residuals   31 0.04742 0.001530

For Batches with more than 175 we can’t reject the Null-hypothesis.

## [1] "P-Values"
## [1] 0.503 0.047 0.007 0.010 0.051 0.000

You can see above the p-values for the pairs (AE > 35): 1 vs. E, 1 vs. F, 1 vs. S, E vs. F, E vs. S and F vs S.

We have to reject (>90% significance level) all Nullhypotheses (batches with different series are same severe) except for the Series 1 vs. E.

Note: Same Analysis with different types of AE (10, 25, 50, 100, 175) come to similar p-values.

## [1] "EJ6796" "EM0477" "EX3510" "EX3599"
## [1] "P-Values"
## [1] 0.739 0.000 0.001 0.000 0.001 0.016

As you can see above all 4 Batches from the “E” series have > 300 AE are in an age range of 52 - 59 and are very different in severity (as also lethality). You can see from the p-values above, that only EJ6796 vs. EM0477 can’t reject the Nullhypothesis. EJ6796 vs. EX3510, EJ6796 vs. EX3599, EM0477 vs. EX3510, EM0477 vs. EX3599 and also EX3510 vs. EX3599 must reject the Hypothesis that these batches are equally same severe by a strong significance level of >98%.

ANOVA:

Conclusion: Empirical testing has to reject the null-hypothesis (H0: All batches are the same) and it seems that the theoretical explanations with variation of active mRNA could be the reason why (https://www.bitchute.com/video/8tbOPuO0jyuV/).

Correlation isn’t causation

So there is a correlation between batches and severity. Now we all know that correlation isn’t causation. But we know that correlation is necessary for causation. I don’t want to dig deeper into this topic but I just want to point out to the Bradford-Hill Criteria:

  1. Strength: The bigger the effect the more likely that it’s causal
  2. Consistency: If different observers in different places (countries) with different samples come to similar conclusions a causation is more likely (https://www.transparenztest.de/post/pei-bericht-244576-covid-impf-nebenwirkungen-und-2255-todesfaelle, https://twitter.com/theotherphilipp/status/1486030456810356737, https://covid-crime.org/ema-numbers-12-21/)
  3. Specificity: Causation is likely if there is a very specific population at a specific site and disease with no other likely explanation
  4. Temporality: The effect has to occur after the cause - e.g. if it is more likely due to covid why is the correlation not in 2020?
  5. Biological gradient: Greater exposure should generally lead to greater incidence of the effect (https://howbad.info/sweden.html)
  6. Plausibility: A plausible mechanism between cause and effect is helpful (https://www.nejm.org/doi/suppl/10.1056/NEJMoa2113017/suppl_file/nejmoa2113017_appendix.pdf)

7.Coherence: Coherence between epidemiological and laboratory findings increases the likelihood of an effect

8.Experiment: Occasionally it’s possible to appeal to experimental evidence FDA

  1. Analogy: The effect of similar factors may be considered FDA


Is there a correlation between age and severity/lethality?

For the german data in VAERS there is an approx. 50% bias in the age-column because in every 2. record they don’t fill the age year but in most cases (97%) they point it out in the symptoms text. After analyzing and correcting this field it seems that there is an underreporting of ~50% over all age-groups as you can see in the next chart. So my assumption from my last update: “At this moment we just can assume that this happens randomly and it will not skew the data too much.” is ok.

So maybe the difference in severity is not because of difference in active mRNA (toxicity) but because of age-dependence. One could argue that some batches were just given to specific age-groups for example in care centres. So if this would be the case we should at least need to see a correlation between severity/lethality and the average age per batch. At first I want to give you the summary of the average age which reported adverse events (AE > 10).

##     AVG_AGE     
##  Min.   :28.12  
##  1st Qu.:42.13  
##  Median :47.51  
##  Mean   :48.41  
##  3rd Qu.:54.78  
##  Max.   :73.43

As we can see in the plots above all show medium strength. In the upper two plots there is (0.318 - 0.361) between average age and severity. In the more robust plots beneath is (0.326 - 0.428). This shows that age has a bit of an explanation but can’t explain or predict the values very well.

## 
##  Pearson's product-moment correlation
## 
## data:  x and y
## t = 4.5611, df = 139, p-value = 1.107e-05
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  0.2078945 0.4965044
## sample estimates:
##       cor 
## 0.3608066

As you can see above the p-value for the correlation test for AE > 11 (upper right plot) is highly significant. This means the correlation of 0.361 is highly significant. So is there really a strong correlation? For a correlation test there is the assumption that there is an underlying normal distribution:

## 
##  Shapiro-Wilk normality test
## 
## data:  Age_DB$AVG_AGE
## W = 0.98959, p-value = 0.3767
## 
##  Shapiro-Wilk normality test
## 
## data:  Age_DB$SEVERITY
## W = 0.95075, p-value = 6.531e-05

As you can see above while this is true for Avg_Age (High p-value - Shapiro tests H0 for Normal-Distribution) this is highly untrue for severity. This means that our data is skewed and the interpretation of the correlation value is less valid with the cor.test.

To fix this Problem I will run a Correlation Coefficient Permutation Test which don’t need this assumption.

Even after calculating the r with permutation test the correlation coefficient seems to be highly significant.

Nevertheless right now as you can see in the plot with the four charts the R² are relatively low which seems that AVG_AGE has a bit of explanation to the severity variable but this model explains nearly nothing especially why are there so many outliers (in both ways) around the age of 50 to 60?

Further more we can see the following:

Note: As you can see in the table on the beginning of this chapter mean and median is approximately the same value for the age variable so there is not a bias because of less robustness against outlier by mean.

For batches with more than 175 reports the ones with the highest severity rate are almost older than younger. But you can see also two strong outliers (FF0900 and 1D020A). The first has an average age of 39 and the second one of 42. And even the most severe batches have an average age around 58 not 88.

So high severity has something to do with age but not as much and not as good as we had assumed. Therefore it is an variable for explanation but without more independet variables the model is not very good for prediciton or explanation. You could also use just the mean of severity and it would likely be not really more bad than this model.

Here is the Lethality Chart filled by average age of deaths:

We can see that, the majority of lower outlier batches by lethality are younger than older et vice versa. But There are also two outliers with very low lethality, relative high AE and an average age of around 83 - 84 (FE7011 & ET3674).

Now I plotted the average age by death and Lethality:

While average age is a weak to moderate good predictor for death per batch (r: 0.44) it seems like lethality (Death/AE) has a clear linear connection (r: 0.65) and the logarithm-model is even better (r: 0.72).

Interim conclusion:

It seems that average (or median) age has a lot of explanation power for lethality but less for death. It could be that this is the case because the absolute numbers of death are batch-size related. As you can see in the right chart there is an exponential trend. In fact the logarithm-model is the appropriate model. But even the logarithm-model can’t explain all of the variance e.g. why there are some low lethality batches and some high lethality batches in the same age group. There could be also differences in gender, health status, comorbidity, severity of injection or mRNA integrity. We therefore need a multiple variable model. Which I will analyze in the chapter after next (Death Modelling). And keep in mind that the median age of death in Germany is 83 (Risk of Disease) which means the life expectancy are normally 10 years higher.

While average age seems to has a lot of explanation power for lethality it’s less good for severity (single and multiple measure). The unexplained variance is far bigger and furthermore a log-model on severity makes the R² worse not better. So for severity we need more and especially better predictor than average age.

It seems like the injection damages the younger ones, damages more heavily the older ones and kills the frail elderly.

Is AE a good predictor for severity?

As seen in the chart AE is pretty bad predictor for severity. So we can’t say: Batches with high Adverse Event reports a more severe. There is no correlation. Therefore I advise to focus generally more on severity than AE when looking at a batch.

Time to death

Normal expected Curve

Stacked Death in Germany:

StackedDeath

The Chart above shows the stacked death over the last 3 Years (whole Years) over the timespan of a year. It includes a year without “pandemic” and “vaccine” (2019), a year with “pandamic” and the most of the year no “vaccine” (2020) and a year with “pandemic” and “vaccine” (2021).

If the chart is shown cumulated you can get the expected deathcurve for germany over time.

Deathcurve

Deathcurves for Covid

The following Charts show the Deaths in correlation to Covid per Week: (https://www.rki.de/DE/Content/InfAZ/N/Neuartiges_Coronavirus/Projekte_RKI/COVID-19_Todesfaelle.html)

Deathcurve

Deathcurve

Now we compare the Normal expected Curves with the Covid-Curves:

Deathcurve

Deathcurve

As shown in the two charts above the Death related to Covid didn’t really influenced the normal overall deathcurve.

Deathcurve after VAX_DATE

As it is shown in the chart above the majority of deaths occured within the first 2 weeks after the vaccination. The gini-coefficient shows how strong the inequality is. If the vaccine has no influence on death at all we would rather see a curve like the expected (diagonally) line.

Death after VAX (SYMPTOMS)

The next Table shows the count of symptoms in correlation to the death after vaccination overall.

The next Table shows the count of symptoms in correlation to the death for different periods.

Death Modelling

Lethality

As we saw was average age a predictor with high explanatory power for lethality but with low for severity.

Now I will adjust our one variable model (average age) to a multiple variable model. Potential variables for lethality could be derived from a predictive model and risk factors for case fatality of covid-19 - a potential model which I could imagine is approx. used by life insurance companies, too:

Death_Covid_by_Age_and_Sex

  • AVG_AGE
  • SEX
  • SEVERITY
    • Single SEVERE / AE
    • Multiple SEVERE/ AE - (L_LTREAHT + DISABLE)/AE
  • Comorbidity
  • mRNA Integrity (I don’t have the german data so this is a variable for the US Data-Model)
## 
## Call:
## lm(formula = log(DIED) ~ AVG_AGE + SEX + MULTIPLE_SEVERITY + 
##     COMORBIDITY_RATE + Median_Days, data = database_multiple_model)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -0.87525 -0.29628  0.00019  0.39957  1.35297 
## 
## Coefficients:
##                    Estimate Std. Error t value Pr(>|t|)   
## (Intercept)        0.059416   0.905947   0.066  0.94817   
## AVG_AGE            0.031561   0.009534   3.310  0.00257 **
## SEX               -0.625953   0.699711  -0.895  0.37863   
## MULTIPLE_SEVERITY  8.150025   4.777979   1.706  0.09913 . 
## COMORBIDITY_RATE   2.775585   3.550056   0.782  0.44087   
## Median_Days       -0.030765   0.012343  -2.492  0.01887 * 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.5789 on 28 degrees of freedom
## Multiple R-squared:  0.503,  Adjusted R-squared:  0.4143 
## F-statistic: 5.669 on 5 and 28 DF,  p-value: 0.0009865
## 
## Call:
## lm(formula = log(LETHALITY) ~ AVG_AGE + SEX + MULTIPLE_SEVERITY + 
##     COMORBIDITY_RATE + Median_Days, data = database_multiple_model)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -0.91162 -0.19011  0.04805  0.18963  0.94097 
## 
## Coefficients:
##                    Estimate Std. Error t value Pr(>|t|)    
## (Intercept)       -6.730721   0.641729 -10.488 3.32e-11 ***
## AVG_AGE            0.039403   0.006753   5.835 2.86e-06 ***
## SEX                0.012569   0.495642   0.025   0.9799    
## MULTIPLE_SEVERITY  8.250196   3.384489   2.438   0.0214 *  
## COMORBIDITY_RATE   0.848463   2.514688   0.337   0.7383    
## Median_Days       -0.008115   0.008743  -0.928   0.3613    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.4101 on 28 degrees of freedom
## Multiple R-squared:  0.6281, Adjusted R-squared:  0.5617 
## F-statistic:  9.46 on 5 and 28 DF,  p-value: 2.285e-05

Note: Model with single severity is worse than multiple. I don’t summary this model here.

As we can see in the output above SEX and COMORBIDITY_RATE (based on at least one current illness) is a pretty bad predictor for lethality. COMORBIDITY is calculated by the field CUR_ILL with TRUE or FALSE, there is no multiple comorbidity accounted yet because the field is relatively bad filled.COMORBIDITY_RATE is in relation to AE. To get a more accurate picture of comorbidity I think it would be necessary to weigh them. Right now we have to assume that SEX and COMORBIDITY_RATE don’t help to explain the lethality. Because of this I get rid of it in the next step:

Short comparison of death by SEX

As you can see on Lot/Batches Searching for Germany (AE > 10)) out of 145 Batches there are 46 (31.7%) batches that have more than 50% males per batch, 6 (4.1%) have the same amount of men and women and 93 (64.2%) have more than 50% females per batch. But we can’t see in the following charts significant excess death in women overall.

The following chart shows the death for female over the last 5 years in Germany:

Death_Female

The following chart shows the death for Male over the last 5 years in Germany:

Death_Male

The following chart shows the death in comparsion for Sex and Year over the last 5 years in Germany:

Death_Female_and_Male

As you can see above the chart shows that it’s more likely to be a female in the majority of the batches with AE > 175. But there is no clear pattern like more women more damage. It’s more like the majority is bulking around 0.5 - 0.7 with a variance in severity by a factor up to 2 times.

Short comparison of death by AGE

Death_45

Death_60

Death_80

Death_80+

## 
## Call:
## lm(formula = log(LETHALITY) ~ AVG_AGE + MULTIPLE_SEVERITY, data = database_multiple_model)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -1.02020 -0.18512  0.03003  0.16771  0.91812 
## 
## Coefficients:
##                   Estimate Std. Error t value Pr(>|t|)    
## (Intercept)       -6.88715    0.53190 -12.948 4.82e-14 ***
## AVG_AGE            0.04110    0.00605   6.794 1.31e-07 ***
## MULTIPLE_SEVERITY  8.49287    3.22980   2.630   0.0132 *  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.3962 on 31 degrees of freedom
## Multiple R-squared:  0.6156, Adjusted R-squared:  0.5908 
## F-statistic: 24.82 on 2 and 31 DF,  p-value: 3.663e-07

While we can see that average age is the dominant predictor, we can also see that multiple severity is also a significant predictor. Therefore we could increase the R² from log-model only age: 0.523 to 0.591.

Severity
## 
## Call:
## lm(formula = SEVERITY ~ AVG_AGE + SEX + COMORBIDITY_RATE + Median_Days, 
##     data = database_multiple_model)
## 
## Residuals:
##       Min        1Q    Median        3Q       Max 
## -0.084163 -0.020311 -0.004876  0.019895  0.071397 
## 
## Coefficients:
##                    Estimate Std. Error t value Pr(>|t|)  
## (Intercept)       0.0322123  0.0487758   0.660   0.5142  
## AVG_AGE           0.0015049  0.0006070   2.479   0.0192 *
## SEX               0.0116171  0.0447775   0.259   0.7971  
## COMORBIDITY_RATE -0.1739572  0.2261323  -0.769   0.4480  
## Median_Days      -0.0005447  0.0007859  -0.693   0.4938  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.03705 on 29 degrees of freedom
## Multiple R-squared:  0.2229, Adjusted R-squared:  0.1157 
## F-statistic:  2.08 on 4 and 29 DF,  p-value: 0.1092

As it’s shown in the table above only average age has a low significant explanation power to severity. Therefore the multiple model don’t help to explain more variance.

As I pointed out in the lethality chapter maybe “just” the count of comorbidity can’t explain variance but maybe the severity of comorbidity. I think this is something to further investigate.

Another really important factor could be the %mRNA Integrity to get a higher explanation rate. Sasha Latypova found with a normal linear model that %mRNA Integrity could have approx. the same amount of explanation rate as average age. This must be analyzed next.

Alexandra (Sasha) Latypova

Differences between Countries

This report was focused on germany but there are differences also between countries (https://howbad.info/international.html).

VAERS is a public passive surveillance system

I often read that VAERS is not a valid Database because theoretically every private person could make a report. But after studying the records and after some Text-Mining-Analysis I have to say that at least 80% of all records are highly valid by physicans, healthcare workes or similiar persons. At least 85% of all records come from PEI (Paul-Ehrlich-Institut). Also I pointed out in number 2 of the Bradford-Hill Criteria VAERS isn’t the only database which seems to have similar patterns. PEI (Paul-Ehrlich-Institut), EMA (European Medicines Agency) via EudraVigilance, WHO via Vigiaccess, InEK (Institut für das Entgeltsystem im Krankenhaus) and DoD (Department of Defense) all seems to have similiar pattern about strongly rising Adverse Events more than what should be expected. I think it’s absolutely necessary to investigate further more! Especially with not only passive surveillance systems but also with active ones.

Conclusion

Risk is everywhere. Everyday we are confronted with different types of risks. Risks to fall down the stairs, risk to have a car accident, risk of getting robbed, risks of a getting sick by a respiratory virus and risk of getting adverse events from an injection. There is no one fits all strategy for everyone. Everyone has to choose his own risk appetite. But what is necessary is that everyone is as good informed as he/she can. Only then it’s possible to make a rational decision. This means not that the decision will alaways have the wished outcome, but in the long run the probability is on your side. What we are seeing with Covid is multidimensional. Some got pretty sick and some will maybe become pretty sick in the future. But what we can see is that most people with good immunity system have no big risks to take especially the more younger you are. If you are younger than 30 it’s more likely that you will die from a hornet, wasp or bee sting than from Covid. If you are younger than 70 there is nearly the same chance by dying from drowning than from Covid and even if you are 80+ the chance for dying with Covid is just a little bit higher than from dying in a motor-vehicle-crash. Now you know some of your risk profile. Does it mean that children shouldn’t play in the garden or are you afraid to go swimming in the holidays or do you stay at home instead of driving to your familiy on christmas? In particular with the Covid-injections there seem many things really strange. Normally a Vaccine needs approx. 10 years to be ready to vaccinate many people. The mRNA gen-therapies just needed 1! It seems that they had to cut corners to be so fast. And this is something about some people blows the whistle (e.g. https://twitter.com/IamBrookJackson/with_replies). As you saw Pfizer also not reached the endpoint on the clinicial trial phase 3 (more people died in the mRNA group than in the placebo) and tried to publish their data over a timeframe of 75 years. I think this isn’t something trustworthy. Further more there are strange things happening in the insurance and funeral companies (https://gettr.com/post/pu46g39a32) as also the things Project Veritas is uncovering (https://www.youtube.com/watch?v=6nSXHrmOy8o&t=103s). On the other hand I showed you some analysis of the Adverse Events which occurs in germany but got reported to VAERS. It’s pretty clear that something is strange about the high differences between/within manufacturer and batches. We can’t say the vaccine are the causality for this at this moment. We just can say there are hints for that and that this must be investigated.

“Don’t risk something you have and need for something you don’t have and don’t need!”

Lot/Batches Searching for Germany (AE > 10)

Sum overall (AE > 10)

##  AVG_AGE       AE HOSPITAL L_THREAT  DISABLE     DIED 
##  7020.13 17628.00  8950.00  1121.00   258.00   821.00

Note: > 0.5 Sex per Batch means higher percentage of females et vice versa.